Re: [PATCH 1/2] wifi: mt76: mt7921: Disable powersaving by default

[Mario Limonciello <mario.limonciello@amd.com>]

On 12/13/2023 06:45, Kalle Valo wrote:
> Mario Limonciello <mario.limonciello@amd.com> writes:
> 
>> Several users have reported awful latency when powersaving is enabled
>> with certain access point combinations.
> 
> What APs are these exactly? In the past 802.11 Power Save Mode was
> challenging due to badly behaving APs. But nowadays with so many mobile
> devices in the market I would assume that APs work a lot better. It
> would be best to investigate the issues in detail and try to fix them in
> mt76, assuming the bugs are in mt76 driver or firmware.

In my case I could reproduce the behavior on my Unifi access points. 
I've got a few that devices could roam between.

I also happen to have a laptop on my desk with a WCN6855 that behaves 
just fine with those same APs.

The other people with problems I've asked to come to this thread to 
share more about their devices.

> 
>> It's also reported that the powersaving feature doesn't provide an
>> ample enough savings to justify being enabled by default with these
>> issues.
> 
> Any numbers or how was this concluded?

It was just a data point in the original patch from Sultan (who is on 
CC).  I haven't corroborated it myself.  Once I saw that other kernel 
drivers like iwlwifi are also "defaulting" to off with a module 
parameter I figured it made sense to bring that for discussion.

> 
>> Introduce a module parameter that would control the power saving
>> behavior.  Set it to default as disabled. This mirrors what some other
>> WLAN drivers like iwlwifi do.
> 
> We have already several ways to control 802.11 power save mode:
> 
> * NL80211_CMD_SET_POWER_SAVE (for example used by 'iw set power_save')

I'll experiment with an unpatched kernel just undoing this from here to 
see if it alone improves things.

> 
> * CONFIG_CFG80211_DEFAULT_PS (for kernel level default)
> 
> * WIPHY_FLAG_PS_ON_BY_DEFAULT (for the driver to control the default setting)
> 
> Adding module parameters as a fourth method sounds confusing so not
> really a fan of this. And the bar is quite high for adding new module
> parameters anyway.
> 

Should we also discuss removing the iwlwifi one then too perhaps?
Seems incongruent to offer it for some drivers but not others.